And now for the science bit!

I know not everyone is interested in learning theory or operant conditioning and if you're not, then just turn to another page now! 

Many are aware of Ivan Pavlov and his experiments with salivating dogs but not so many are aware that it is actually much more complicated than that! For those who haven't heard of our friend Pavlov and would like to know more here's a bit of history.  

Ivan Pavlov- 1849-1936

A Russian psychologist studied learning in laboratory dogs who discovered that dogs would salivate after hearing the bell that was rung shortly before food was delivered. He was the first to realise that it was possible for such "classical" conditioning to occur. Basically all this means is that it is possible for an association to take place between an unconditioned stimulus (bell) with the unconditioned stimulus (food) which gives rise to a biological response (salivation). It is on the basis of associative learning that clicker training was developed. 

Many traditional dog trainers work on the theory of association and this is the code of dog training. A dog associates a stimulus with a good event or with a bad event (punishment). However, it is very much more complicated than that.

To keep it very simple (as it has just taken me the last year to learn and digest all this) there are other very important factors involved in how an animal learns. 

Context: The environment is important as learning is accelerated when the same context is present. This is one reason why I see so many people chasing their dogs cross the heath complaining that their dog has the gold good citizen's medal but won't come back outdoors! So the goal is to practise everything in lots of different locations so that the dog can generalise the training given.

Blocking: Something new is not learned because of a prior association with the task of an inhibitory or irrelevant nature. I call this "learned irrelevance". If you are teaching your dog to sit and keep repeating the command "sit" to them without them actually sitting and earning the reward, the word "sit" becomes irrelevant and they continue to fail to respond to that command.

Reward relevance. If the reward is of no interest to the animal it will not respond in connection with it. 

Overshadowing; When two activities are occurring simultaneously and one of them is of greater relevance or surprise to the other. In this case, the stimulus of greater relevance/surprise will be learned more readily and the other ignored. E.G. you are trying to recall your dog for a treat or a toy and another dog turns up and your dog runs off to play with it! Best to teach recall in a quiet environment first so he can learn without interruption. 

Stimulus-stimulus-response. This is where even though a behaviour or task may be expected not to be performed (e.g. an aversive) the animal still does it. The reasons for this are not clear but it does appear that certain behaviours will be performed readily even in the knowledge that the result is an aversive. This is a major weakness of the more traditional training techniques as these often rely on punishment which very often does not work.  

Extinction burst: When trying to modify behaviour and remove rewards for a behaviour which has been previously rewarded the animal strives even harder at first before giving in. I think most of us have seen this in children's behaviour when saying "No" to the sweets at the checkout when in the past they have been accustomed to getting this. The tantrum can be quite amazing before they finally give up! 

Spontaneous recovery. This is a tricky one. When behaviour has been modified, the animal relapses into the old behaviour again some time after the initial extinction burst. This appears to happen because for every behaviour a trace is is forged inthe brain. If the behaviour is repeated often, the trace becomes deeper and more likely to be visited. Rehabilitation aims to create a new behaviour and a new trace. However, the old trace never goes and can be re visited any time (a bit like a dried out river channel,  when it rains, water runs down the old channel creating a new inlet readily). This  often leads people to assume the training/ behavioural techniques do not work. In fact they do, this is just a normal part of recovery and there is nothing you cn do about it other than keep reinforcing the new behaviour in the hope that that trace becomes deeper over time than the old one.

Rescorla - Wagner model (1974)

This is a mathematical model which in its very basic terms means that for learning to take place most effectively the stimulus and the unconditioned stimulus should be both surprising and noticeable. This model depicts a bell shaped learning curve showing a decline in learning as an animal becomes accustomed to a learning task. It is hypothesised that this is because the stimuli are less salient and surprising each learning trial. This is very useful as changes in the rate of learning may be possible by changing certain aspects of the learning trial to increase surprise and salience thus increasing the learning rate. 

Reward schedule: Be consistent. If a behaviour is always reinforced it will continue but without any great passion. If a behaviour is never rewarded it becomes extinct, however, if it is randomly rewarded the behaviour becomes very strong and resistant to extinction. This works for bad behaviour/training as well as good!          

Memory retention: Dogs learn best when the reward is delivered within a couple of seconds of the behaviour which is being reinforced. Studies have shown animals to be able to respond to delayed matching to sample. Tests have shown that animals can learn that rewards correspond to certain behaviours up to 30 seconds apart, however, this was for non human primates and it is generally considered that dogs need to be rewarded within 2 seconds for adequate learning to take place.

Forgetting can occur when an animal is overloaded. Also memory can be influenced by anything occurring immediately prior to or immediately after an event. E.G a classic example might be dog training in a class where a dog keeps barking. Each time dog a) tries to learn task DOWN, dog c) barks, interrupting concentration and slowing down the rate of learning.  

This is a very basic and limited précis of learning theory. The aim of sharing this with you is so that it may be seen that training a dog is not as simple as giving a treat for good behaviour and rattling a can for bad behaviour. As we can see from the above, punishing bad behaviour can increase the behaviour. Punishment also increases fear which inhibits learning and increase aggression. Whilst in a fearful state cortisol is released and it is understood that a part of the brain called the amygdala takes over which inhibits rational thought processes and the ability to learn.  

Learning all this has made me a much better trainer and I now do believe that academic grounding in dog training and behaviour is necessary to help people to train their dogs.  New discoveries are being made all the time and education ensures that you are kept abreast of the most effective techniques available.

If anyone is interested in researching this in much more detail a great book is Pearce, J.M. (2008) Animal Learning and Cognition, An Introduction. Third Edition. New York: Psychology Press.